Will the raw data of WikiKG90M-LSC be released?

[jianyucai]

Hi, I noticed that the raw directory of wikikg90m_kddcup2021 is empty. Will the raw data be released?

Thanks.

[weihua916]

Yes, we will eventually release the raw data, but we have not yet decided the timing for that.

[jianyucai]

Thank you for your kind response.

[hhr114]

Yes, we will eventually release the raw data, but we have not yet decided the timing for that.

Hi, Weihua,
I notice that the KDD Cup 2021 is over. Would u plz tell me when will the raw data be released?

Thanks : )

[zhengfengL]

Hi, I notice that the KDD Cup 2021 is over, when will the raw data be released?

Thanks : )

[weihua916]

Whether we can release the name is still TBD. It’s likely we cannot for the time being since we want to avoid test data leakage.

[cj2001]

I would like to +1 the request to have the labeled data (i.e. the conversion between integers and node/relationship labels) provided. This is a tremendously valuable data set that may in the community can benefit from. Even if it is the deprecated set, it would still be of great benefit to all.

[sanxing-chen]

Whether we can release the name is still TBD. It’s likely we cannot for the time being since we want to avoid test data leakage.

I am curious about why releasing the names of entities would cause test data leakage. @weihua916 could you elaborate on this? Hiding the names prevents the development of inductive models with built-in text encoders.

[weihua916]

It's because you can use the names to search over the WikiData (e.g., Japan: https://www.wikidata.org/wiki/Q17). For now, we have already encoded text for you. You can use the provided embedding features.

[sanxing-chen]

Thanks for the reply, fair enough.

[jannatmeem95]

Hi! Will the raw data be released? Or is there any publicly available small chunk/snippet of the raw dataset?

[weihua916]

Hi everyone, we just released the raw text data: https://groups.google.com/g/open-graph-benchmark/c/R0SKtj9qQyE